This paper addresses the problem of estimation and tracking of a road using the new tilt-rotor convertible MAV (mini aerial vehicle). For that objective, we consider the following scenario: (i) no previous knowledge of the road, i.e. shape, dimension and color, (ii) loss of information by the sensors is considered and (iii) nonlinear dynamics of the MAV is taken into consideration. Aiming at this goal, two operational regions are defined: road detected and road not detected by the sensors. A switching between the measurements of imaging and inertial sensors enables estimation of the required states in both operational regions. For dealing with both aforementioned cases, a Lyapunov-based switching control for stabilizing the vehicle's position is proposed. Unmodeled dynamics such as friction forces are estimated by means of the proposed controller. The global exponential stability of the position subsystem together with the switching controller is demonstrated exploiting the fact that the individual closed-loop systems are globally exponentially stable and the switching is sufficiently slow, so as to allow the transient effects to dissipate after each switch. The control law is validated on the Quad-plane experimental platform, showing the expected behavior during autonomous navigation.